Preserving Inference Privacy in Multi-Agent Systems with Adaptive Modulo Hash Function
Puspanjali Ghoshal, Ashok Singh Sairam · 2025
In a Multi-Agent System, autonomous agents observe the system parameters and send the data to a fusion center. The fusion center aggregates the data to deduce the system parameters with high accuracy. For purposes of authentication, non-repudiation and overall better functioning of the system, the data sent by the agent is appended with its private parameters. However, inclusion of these private parameters coupled with the agent observation leads to inference privacy breaches like profiling. To protect the agents against such breaches, a sanitization function needs to be incorporated at the agent level. One of the many ways to provide inference privacy is via dimension reduction. In this paper, we reduce the dimension of the data tuple by the incorporation of a modulo hash based privacy function. Although hash functions are primarily used for verification of data integrity, we have proposed a novel modulo hash based algorithm for sanitization. Our proposed hash function takes into consideration the level of privacy required by the agent, and is thus tunable.